Github user liancheng commented on a diff in the pull request:
https://github.com/apache/spark/pull/13893#discussion_r73156772
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/planning/patterns.scala
---
@@ -64,10 +64,17 @@ object PhysicalOperation extends PredicateHelper {
val substitutedFields =
fields.map(substitute(aliases)).asInstanceOf[Seq[NamedExpression]]
(Some(substitutedFields), filters, other,
collectAliases(substitutedFields))
- case Filter(condition, child) if condition.deterministic =>
+ case Filter(condition, child) =>
val (fields, filters, other, aliases) =
collectProjectsAndFilters(child)
- val substitutedCondition = substitute(aliases)(condition)
- (fields, filters ++
splitConjunctivePredicates(substitutedCondition), other, aliases)
+
+ // Deterministic parts of filter condition placed before
non-deterministic predicates could
+ // be pushed down safely.
+ val (pushDown, rest) =
--- End diff --
I also think that silently dropping nondeterministic filters can be
dangerous. Maybe we should just return all operators beneath the top-most
nondeterministic filter as the bottom operator?
For example, say we have a plan tree like this:
```
Project a, b
Filter a > 1
Filter b < 3
Filter RAND(42) > 0.5
Filter c < 2
TableScan t
```
We should return the following result:
```
(
// Project list
Seq(a, b),
// Deterministic filters
Seq(b < 3, a > 1),
// The top-most nondeterministic operator filter with all operators
beneath
Filter(
RAND(42) > 0.5,
Filter(c < 2,
TableScan(t)))
)
```
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
